Creating an Open Knowledge Graph for Climate
Day 1 | 10:30 | 00:25 | AW1.126 | Peter Murray-Rust
Note: I'm reworking this at the moment, some things won't work.
semanticClimate is a global hybrid community where interns from colleges (mainly in India) create climate knowledge to help the world make informed decisions. We work with trustable material such as the UN/IPCC reports (over 15,000 pages of important, but dense text). The resulting knowledge products include:
- term-based dictionaries (ontologies) enhanced with Wikipedia and Wikidata
- a Corpus tool for scraping and analysing the current Open scholarly literature
- a knowledge graph created from the above with navigation tools
and F/OSS software to make this easy and automatic.
semanticClimate interns come from high-school up and need have no knowledge of software. They learn-by-doing, and in some weeks have 2-hour online sessions daily - these are recorded and transcribed to text for all to see. Interns are encouraged to give public talks (e.g. OKFN, Wikipedia, CODATA) and to make 5 min videos. All software is modular, Git-branched, versioned and unit-tested. Where possible we publish it in J. Open Source Software.
The session image is part of our Climate Knowledge Graph (my email is [email protected]. I can't change it in the form!)
Links
- Collection of resources for semanticClimate. These include talks, videos and other outreach
- pygetpapers (search scholarly literature) Open Source. (Ayush Garg, see also publication in J. OpenSource Software https://joss.theoj.org/papers/10.21105/joss.04451. AG was still at high school.
- docanalysis (text analysis) Shweata Hegde. SNH was an undergraduate, initially with no software experience
- amilib (Peter Murray-Rust) Python library of > 1000 methods and > 300 user-oriented tests for downloading/transforming/republishing documents
- Semantic Climate Glossary (475 pages) created automatically and enhanced by manual multilingual terms
- Presentation of semanticClimate at OKFN's "The Tech we Want" by Shweata N Hegde (ca 6 mins)